YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
NeuTTS-Nano hi_translit Checkpoint
Fine-tuned on hi_translit speech data for text-to-speech synthesis with voice cloning.
Datasets
data/encoded_hi_indictts
Training
| Param | Value |
|---|---|
| Base model | neuphonic/neutts-nano |
| Max steps | 4000 |
| Learning rate | 5e-05 |
| Batch size | 1 |
| Precision | bf16 |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("vanshnawander/neutts-nano-hi-translit", subfolder="final")
tokenizer = AutoTokenizer.from_pretrained("vanshnawander/neutts-nano-hi-translit", subfolder="final")
Structure
best/— lowest eval loss checkpoint (always kept)final/— last training step (always kept)checkpoint-<step>/— rolling window of last 3 checkpoints
Auto-generated by HFSyncCallback — 2026-02-16 12:53
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support